Bulletin of the World Health Organization

Comparative impact assessment of child pneumonia interventions

Louis Niessen a, Anne ten Hove b, Henk Hilderink c, Martin Weber d, Kim Mulholland e & Majid Ezzati f

a. Department of International Health, Johns Hopkins Bloomberg School of Public Health, 615 N Wolfe Street, Baltimore, MD, 21205, United States of America (USA).
b. Department for Health Policy and Management, Erasmus University, Rotterdam, the Netherlands.
c. Netherlands Environmental Assessment Agency, Bilthoven, the Netherlands.
d. World Health Organization Country Office, Jakarta, Indonesia.
e. Infectious Disease Epidemiology Unit, London School of Hygiene, London, England.
f. Department of Population and International Health and Department of Environmental Health, Harvard School of Public Health, Cambridge, MA, USA.

Correspondence to LW Niessen (e-mail: lniessen@jhsph.edu).

(Submitted: 24 January 2008 – Revised version received: 01 October 2008 – Accepted: 02 October 2008 – Published online: 16 April 2009.)

Bulletin of the World Health Organization 2009;87:472-480. doi: 10.2471/BLT.08.050872


Progress in reducing mortality from pneumonia in children under 5 years of age has been relatively slow in many parts of the developing world, where about 155 million clinical pneumonia episodes and 2 million deaths occur annually.1,2 Risk factors for pneumonia include stunting and underweight,1,3,4 suboptimal breastfeeding,5,6 lack of immunization7,8 and indoor air pollution from household use of solid fuels.912 There is evidence that effective and appropriate management of clinical cases is possible13,14 at health-care facilities15 and in the community,16 but this level of management is often lacking.

Efforts to control pneumonia are needed to meet Millennium Development Goal 4 (MDG 4), to reduce child mortality in the world by two-thirds by 2015.17 Often, a package of priority interventions is developed to address MDG targets and reduce child mortality.4,6,1820 Cost-effectiveness analysis has become vital in deciding what interventions to implement and scale up.21 Single-candidate interventions to reduce pneumonia have been evaluated in general economic terms,6,11,18,2224 but no comprehensive analysis has focused on pneumonia control.

Different interventions can affect incidence or case fatality, with differences noted across age groups. Population risk interventions can target specific subpopulations, while immunization is intended for all infants. Preventive interventions of this kind may reduce the incidence of pneumonia, whereas case management influences case fatality after falling ill. Both types of interventions can reduce pneumonia mortality.

The aim of this study was to compare the impact of eight preventive and curative interventions at the population level6,2527 and to identify the intervention mixes that generate the highest possible level of child health at the lowest cost.


To estimate the population health effects and total costs of pneumonia interventions from a health-care perspective, we applied demographic life tables for the 40 countries with the highest mortality (list available at: http://oldwww.bmg.eur.nl/personal/niessen/Webtable%20Countries%20by%20Region.doc The tables were used to estimate the health effect of risk factors, as well as the reductions in incidence and case fatality in population cohorts, simultaneously and consistently.6,2527 Detailed descriptions of concepts, methods, background papers, regional studies and data are available at: WHO-CHOICE (CHOosing Interventions that are Cost Effective) at: www.who.int/choice/en. Box 1 provides an overview of the approach.

Box 1. Stepwise description of impact assessment for comparatively analysing the costs and effects of interventions6,25,26

1. Construct epidemiologic disease model. Give a population-based description; establish how parameters of the disease model interact (i.e. relative risks, incidence, case-fatality ratio, neonatal and mortality rates, by age group and sex);

2. Review national data for year of study. Include population structure and absolute figures, births, pneumonia epidemiologic rates and intervention coverage;

3. Construct baseline epidemiological parameters. Reflect current population figures and epidemiologic rates, a situation of limited health care and the future United Nations demographic scenario;

4. Estimate effectiveness. Repeat analysis under Step 3 with changes to one or more key epidemiological parameters (incidence or case–fatality rate) as a result of intervention effectiveness; compute the total number of healthy life years gained (or of DALYs averted).

5. Estimate costs. Establish coverage and contact rates; apply unit costs and add programme costs by intervention mix.

6. Generate a cost and effectiveness league table. Estimate the total costs and health benefits (DALYs averted) of single interventions and intervention mixes and establish a ranking table based on the cost-effectiveness ratio.

DALY, disability-adjusted life year.

We considered the epidemiological characteristics and level of health care of each of the 40 countries, as well as the coverage levels of the expanded programme on immunization (EPI) and of facility-based case management. Due to the large uncertainties involved in the epidemiologic, effectiveness and cost estimates, we included a high and a low cost-effectiveness scenario for each mix of interventions.

Each country’s life table provides summary estimates of how pneumonia affects mortality and morbidity, expressed in terms of disability-adjusted life years (DALYs) lost. The tables also provide estimates of the effect and cost of mixed health interventions, in United States dollars (US$) for the year 2000, with a 3% discount rate according to health economics guidelines. We combined estimated health gains and costs per intervention to identify the sets of health interventions that maximized child health at different budget levels by providing the greatest health yield per dollar spent. The life tables were implemented in C++ (a general programming language) using M language (a language for working with data and building domain models). The script with M-equations is available at: http://oldwww.bmg.eur.nl/persona/niessen/GAPP_LOW.MPdf.pdf

Epidemiologic and demographic data

The life tables used in the model were based on the recently published WHO country data, which draw on reviews of incidence and mortality for childhood and neonatal pneumonia. Incidence estimates were taken from the epidemiological review.1 Consistent applied case-fatality rates were calculated by dividing annual incidence figures by annual mortality rates from the global burden of disease data set.2,7

Risk factor prevalence data were derived from the WHO Statistical Information System (WHOSIS), available at: http://www.who.int/whosis/en/index.html They included nonexclusive breastfeeding, undernutrition (defined as underweight for age, z < −2), measles immunization coverage and exposure to indoor air pollution in the population under 5 years of age. Relative risks of indoor air pollution by type of biomass fuel for pneumonia incidence were derived from the Global Action Plan for Pneumonia (GAPP).9,10,28 Other relative risks for pneumonia incidence were obtained from the same review.1

National statistics on neonatal, infant and child mortality for 2005 were obtained from the online database of the Institute for Health Metrics and Evaluation.7 The fractions of neonatal mortality attributable to pneumonia and sepsis were obtained from the The Lancet nutrition series.20 Case-fatality ratios for children are specified by three age groups: neonatal period until 1 month of age, remainder of the first year (2−12 months of age) and 1–5 years.

The disability weight used to compute pneumonia morbidity for a disease episode lasting 2 weeks was 0.279.29 DALYs were calculated by applying the region-specific disability weights for the general population by age and sex.30 Country-level demographic data on population structure, birth rates and general mortality rates were obtained from official 2005 estimates by the Population Reference Bureau (available at: http://www.prb.org/Publications/Datasheets/2008/2008wpds.aspx).

Interventions, effectiveness and costs

In all scenarios we assumed a programme effectiveness time horizon of 10 years for all interventions, starting in 2005. After that, the new population cohorts resumed pre-intervention status, in line with the standardized cost-effectiveness approach of WHO-CHOICE project.6 The calculations included the extra life-years lived by additional surviving children beyond the 10-year period, as well as the pneumonia incidence reduction from immunization until the last immunized age group reaches the age of 5 years (in 2020). We estimated total health effects over a period of 100 years to include all life-years gained beyond the 10-year time horizon, among all survivors. We calculated intervention costs in International dollars (I$) to allow comparisons.

Table 1 shows the selected interventions and related input data for various scenarios. The subsections below describe the scenario assumptions by intervention category.

Reduction of indoor air pollution

The 90% confidence interval (CI) of the relative risk (RR) of pneumonia due to exposure to indoor air pollution was estimated to be 1.42 to 2.53.28 Two interventions for indoor air pollution were selected.9,28 The first was a switch at the household level to cleaner gaseous or liquid fuels (liquefied petroleum gas, kerosene or ethanol); the second was better combustion ventilation through high-quality and well-maintained biomass stoves. The health effect of intervening against this risk factor derives primarily from observational studies (including one unpublished randomized study of high-quality stoves). The GAPP reviews assumed that introducing cleaner fuel reduces pneumonia risk.9,28 Based on this assumption, changing to full-scale cleaner household fuel could lower pneumonia incidence by 50% (the attributable burden for indoor air pollution). However, high-quality, well-maintained stoves are not expected to prevent all exposure to indoor air pollution. In an earlier review and cost-effectiveness study, a 75% reduction in exposure was assumed in a scenario of full coverage with good stoves.11 Given the high and low RRs linked to indoor air pollution under this scenario (equations in Table 1), the pneumonia incidence reduction would be 22.2% to 45.8%.28 The cost methodology and actual cost estimates are based on WHO reports11,23 with a two- or four-year stove lifetime.

Nutritional interventions

Selected nutritional interventions to reduce pneumonia were exclusive breastfeeding promotion up to 6 months of age6,18 and food supplementation with zinc.3,33 Region-specific cost estimates were based on those from the WHO-CHOICE programme.


The scenarios included two vaccines as potential interventions to reduce pneumonia risk. The measles vaccine was not included, since its already high coverage in most of the 40 countries studied would have made its effect on pneumonia mortality difficult to quantify. The population effectiveness of immunization depends on the level of protection against the bacteria (Hib and pneumococcus), but even more on the actual attributable contribution of these bacteria to the pneumonia burden. Hib and pneumococcus may account for more than half of pneumonia mortality in children.17 The relative importance of these bacteria as causes of pneumonia in different settings is unknown, but the similarity of the trial results suggests that major differences between populations do not exist. The effectiveness range given by the high and low country scenarios takes into account the variety of agents (Table 1). The joint effect of the two vaccination programmes targeting two different microorganisms was assumed to be additive. The cost estimates were based on earlier economic evaluations.22,24 Implementation was assumed to occur within existing immunization programmes and infrastructure.

Pneumonia case management

Two delivery strategies were chosen to treat children with pneumonia: a facility‑based approach,15 and a community-level approach in which children were diagnosed and treated by community health workers.16 The estimated reduction of pneumonia mortality through pneumonia case management was based on two reviews.13,16 These reported an efficacious (i.e. under ideal circumstances) reduction of 42% (90% CI: 22–57) in neonatal pneumonia mortality and of 36% (90% CI: 20–49) in child pneumonia mortality, confirmed by a review of management by community health workers.31 We subtracted these expected reductions from the country-specific, age-specific case fatality rates, while we included the uncertainty range based on the CI. In severe cases (8.6% of all incident cases), we assumed a case fatality reduction of 51%.32 The cost data of case-management strategies at the facility level are from the WHO-CHOICE programme and updates by WHO’s Child and Adolescent Health Department.21,32 The community-based cost estimates are from the Disease Control Priorities in Developing Countries (DCP2) project.14 We varied the number of budgeted visits by a village agent to children treated for pneumonia by one (low-cost scenario) to two times (high-cost scenario).14,31


Table 2 shows the regional aggregate results on the effect of using solid fuels on pneumonia mortality in children. Table 3 shows the potential impact of pneumonia interventions on total mortality among children under 5, and Table 4 lists the cost-effectiveness ratios. In each table, all eight intervention options are grouped into the four intervention areas described above (indoor air pollution, undernutrition, immunization and case management). In the country profiles, further expansion of pneumonia programmes is considered, alongside existing vaccination programmes and curative services.

The attributable pneumonia burden due to indoor air pollution by WHO region was based on the country-specific exposure estimates from the WHOSIS database. The two countries with the largest populations – China and India – showed a high level (> 70%) of solid fuel use. The attributable burden for indoor air pollution in world regions varied from 10% to 38%, with a limited uncertainty range. The contribution of indoor air pollution to the global burden of childhood pneumonia is large (30%; CI: 18–44). Table 3 provides the aggregated results by WHO region of health gains for two intervention packages in the high-burden countries.

Table 5, which illustrates the possibilities for country-level policy-making, presents two country profiles with combinations of eight intervention scenarios. Both single (Table 4) and combined interventions (Table 5) show low-cost outcomes between I$ 10 and I$ 60 per DALY averted for interventions in the WHO Africa D and E subregions, and in the WHO Eastern Mediterranean D subregion. In other regions, effective options were immunization, nutritional interventions and community-based case management. A listing of WHO epidemiological subregions is available at: http://www.who.int/choice/demography/regions Many mixes of interventions fell in the cost range of I$ 60 to I$ 120 per DALY averted; others were less cost-effective in light of the general country income level. In some poorer regions, the two indoor air pollution interventions showed the same cost-effective levels as other interventions. In general, the indoor air pollution interventions appear to be less cost-effective than other interventions for reducing pneumonia mortality. The maximum potential reduction in child mortality, given existing infrastructures and including indoor air pollution interventions, appears to be about 13–17%. Thus, most of the child pneumonia mortality could be avoided if all interventions were implemented.


Population-based preventive measures and expanding community-based case management appear to be the most effective options for reducing pneumonia mortality. Adding these measures to existing facility-based case management would increase the efficiency of health system as a whole. When outreach expansion is limited and infrastructure is lacking, immunization is costly. Where measles vaccination coverage is already high, both types of pneumonia vaccine are attractive options. The estimates on immunization depend strongly on the price per dose. Expanded case management, combined with expanded use of new vaccines, would increase system efficiency further. Adding new vaccines and expanding immunization coverage, nutritional interventions and community case management lead to relatively cost-effective pneumonia packages, as compared with facility-based management alone, because the latter was more costly in all scenarios.

Additionally, we found that health risk reduction through nutritional and immunization intervention programmes increases the cost-effectiveness of programmes for case management of childhood illnesses. The region and country league tables present the additional cost-effective options of expanded community case management and improved neonatal management.

The cost-effectiveness results showed the efficiency of implementing interventions alongside an existing health care structure, in comparison with a baseline situation. Presenting the results in this way provides policy-makers with a general impression of the impact of an intervention; it also makes it possible to compare the efficiency of existing and new packages and possible ways to improve the allocation of funds. For example, in a country such as Guatemala, the most attractive additional options would be zinc supplementation combined with community case management. If these interventions were introduced simultaneously with the available environmental interventions, the additional cost of the package per DALY would increase. When environmental interventions are introduced wherever other interventions are already in place, the extra health benefits are limited and the additional cost per DALY (i.e. marginal cost-effectiveness) can be high. For example, in a country such as Nigeria, which has some infrastructure but no proven options to reduce indoor air pollution, including up-scaling community case management, along with preventive programmes, would increase the cost-effectiveness of implementing a pneumonia control package.

Data are limited in almost all countries. Detailed data on pneumonia deaths are lacking, and community-based data on clinical episodes are sparse.1 Research is needed to better diagnose pneumonia and identify it as the cause of death. Our results are therefore difficult to validate beyond the recent reviews presented, whose quality determines the results of the economic impact evaluation. We were unable to distinguish between studies that reported intervention efficacy and those that reported community effectiveness. We attempted to consider this issue and other sources of uncertainty in our high and low effectiveness and cost scenarios; however, better data on community effectiveness and associated costs are needed. New preventive interventions may lead to net cost savings by preventing costly disease. However, we did not take into account potential savings due to cost offsets, lower use of health services and averted loss of workdays due to fewer illness episodes. Our results are thus conservative.

A point of debate is the cost of investing in cleaner fuels, whose cost per DALY averted is higher than that of other options. The results are not directly comparable, however, because the cost of cleaner fuels is offset by other societal benefits, such as time saved looking for firewood or other biomass fuels. If only the additional implementation efforts in an already existing health sector setting are considered and the extra costs of clean fuels are ignored, the cost-effectiveness ratio is lower. Uncertainty also surrounds the effectiveness and cost of community case management programmes. These are likely to be directly correlated with the quality improvements and the additional cost per village of visits by a village agent. These variables make it difficult to draw definite conclusions from the economic evaluation of these interventions. Still, our studies have identified three potentially valuable interventions to improve child survival: nutritional interventions, immunization and low-cost, effective case management. Innovative use of vaccines, focusing on the highest at-risk groups, could amplify the impact.

National priorities

Donors and national agencies involved in child survival programmes need to select those that maximize child health after considering existing mortality levels, infrastructure and funds available.34 The present study, focused on children, offers policy-makers a range of potential pneumonia interventions and estimates of the money they require.35

Internationally, there is agreement on using disease-burden estimates and data on the cost-effectiveness of interventions to select priority areas. New insights should be applied in real-life country settings to find local solutions and implement appropriate options. Country programme managers need more specific information on the effects and costs of child programmes so they can weigh them against other criteria, such as equity and other societal benefits.3537

We included in our scenarios only interventions for which effectiveness data were available. Due to a lack of data we could not examine the management of severe malnutrition through improved complementary feeding or strong community programmes. Malnutrition is a major risk factor for severe pneumonia,1 yet no adequate study of the preventive effectiveness of such programmes has been performed.

The links between evidence and policy tend to be weak because national policies are the outcome of complicated processes among parties with different interests.36,37 Impact analysis strengthens the selection of optimum child packages, and this paper shows how policy in this area can be more evidence based. ■


We thank the two anonymous reviewers for their detailed comments, and the members of the Global Action Plan for Pneumonia review groups for their scientific contributions. We also thank Shamim Qazi for his coordinating efforts and support.

Funding: The research is supported by a grant from the Netherlands Environmental Assessment Agency on integrated modelling, while two expert workshops were funded by WHO and the United Nations Children’s Fund.

Competing interests: None declared.